WHY AND HOW CULTURE MATTERS IN COMMUNITY INTERORGANIZATIONAL STRUCTURE

Transcription

WHY AND HOW CULTURE MATTERS IN COMMUNITY INTERORGANIZATIONAL STRUCTURE
WHY AND HOW CULTURE MATTERS IN COMMUNITY
INTERORGANIZATIONAL STRUCTURE
MOSES A. BOUDOURIDES
Abstract. Our aim here is to plead for the significance of cultural considerations of overlapping inter–attitudinal patterns right next to well established
structural considerations of interorganizational networks based on overlapping
membership patterns. In particular, we examine how the analytic methodological incorporation of cultural attributes or attitudes might enhance our
understanding of structural community categorizations in interorganizational
networks. For this purpose, we analyze data of the International Peace Protest
Survey (IPPS) on the world–wide peace protests of February, 15, 2003, in order to manifest the added value offered by the consideration of the culture–
structure duality in participation studies.
1. Structure, Culture and Social Networks
In conformity to the “cultural turn” in modern sociology, culture and structure (or
the cultural and the social) are usually conceived intermingled in a complex duality
(Friedland & Mohr, 2004). As a matter of fact, that was also the fundamental conviction of the structural framework in the general sociology of Harrison C. White
(2008a), which, in particular, was preceded by his less known “catnets” thesis aiming to provide a preliminary analytic formalization of the structure–culture duality
(White, 2008b). Similar sources of inspiration were met in the dogma of “relational sociology” elaborated in the famous Manifesto for a Relational Sociology of
Mustafa Emirbayer (1997) and in the much broader discussion on the relationship
between structure and agency that took place in the context of collective action
(Emirbayer & Mische, 1998, Emirbayer & Goodwin, 1994) – although the latter
was overtly critical to White’s approach. In any case, under the connotation of all
the fundamental contributions on structural–cultural relationalism – and beyond
their apparent disparities – the term “cultural turn in social network analysis”
(M¨
utzel, 2009) is sometimes used in the context of certain analytic approaches,
which prefer to treat action–based social networks as both constituting and constituted by culture.
Now, let us recall that the inaugurating idea of social network analysis was that
social structure is not to be understood in terms of categorical attributes and, in
particular, of any cultural identifications of individual or collective actors, but solely
Date: Draft version, April 26, 2011.
Part of this work was done while the author was visiting the Eindhoven University of Technology in the Netherlands from October 2010 to January 2011, the Antwerp Management School in
Belgium during February and March 2011 and the Universit´
e Paris Dauphine in France in April
2011. It is intended to be a first draft of the author’s contribution to a paper to be written with
Iosif Botetzagias and Stefaan Walgrave.
1
2
M.A. BOUDOURIDES
on the basis of social interaction among actors. Nevertheless, this anticategorical
imperative of network analysis (Emirbayer & Goodwin, 1994) should not be construed as if the relational theorization of structure and culture was postulating that
actors, ties and attributes were all independent to each other. On the opposite, it is
in the enmeshing of structure and culture that meaning arises and identities emerge
concurrently and viscerally coupled with materialized patterns of structural rootedness, i.e., embeddedness. For instance, from Harrison C. White’s (2008a) point
of view, actors make sense of themselves and the world they live, while striving to
establish control over surrounding ambiguities and contingencies and to settle their
social footing with other actors or entities in their everyday life.
In particular, when structure is formally represented by a pattern of relations
(or ties) among actors and culture is schematically reflected into a multiplicity
of variable and diverse attributes or attitudes that actors exhibit, there are two
coupled mechanisms and processes of social change and morphogenesis, which are
commonly examined in social network analysis. On the one hand, it is through
the mechanism of network influence that structure may determine culture, in the
sense that actors’ attitudes might change due to the influence exerted from their
network localities. At the same time, on the other hand, it is through the mechanism of network selection that culture may determine structure, in the sense that
culture may motivate actors to establish new connections or severe older ones; in
this way, actors manage to interlace their structural entanglements, while driven
by the cultural attributes they possess as holders of attitudes conforming to norms
or codes of behavior, which are dictated (positively or negatively) by the prevailing cultural dispositions (habitus, preferences, morals etc.). However, these two
mechanisms and the processes they set off are not mutually exclusive, but, rather,
intertwined, as many network theorists have already argued (Erickson, 1988, Leenders, 1997). Thus, for instance, in the cases of homophily (Lazarsfeld & Merton,
1964, McPherson et al. , 2001) or focus constraint (Feld, 2000), cultural attributes
and actors’ attitudes were studied as predictors of structure (e.g., Robins et al. ,
2001b, 2001a). Furthermore, in other formal studies, structural ties were taken as
predictors of cultural attributes and actors’ attitudes (cf., Robins et al. , 2001a;
relevant statistical approaches for both cases are discussed by Robins et al. , 2007,
Steglich et al. , 2007).
2. Interorganizational Networks and Culture
In social network analysis, interorganizational networks are understood as networks,
which are formed by a particular class of collective actors, organizations. In such
networks, ties among actors might be of two types. One is the type of direct interorganizational ties – such as, for instance, on the macro-level, various exchanges
among organizations or, on the micro-level, some sort of formal or informal interactions among members of organizations. Analytically, these networks are commonly
described as one-mode networks, since they involve a single mode of actors (organizations) together with the pattern of ties among them. However, here, we will be
concerned with the other type of indirect (or “latent” in the terminology of Alberto
Melucci, 1989, 1996) ties among organizations, which are commonly postulated to
exist through and because of the occurrence of patterns of multiple–memberships
CULTURE MATTERS IN COMMUNITY INTERORGANIZATIONAL STRUCTURE
3
of individual actors, who simultaneously participate or are involved in more than
one of the studied organizations. From the analytic (graph–theoretic) point of
view, such structures are called two-mode networks (or bipartite graphs), as they
involve two modes of actors, the mode of individuals and the mode of organizations,
together with the pattern of associations (affiliations or memberships) among individuals and organizations – notwithstanding, without assigning any direct linkages
among either one of the actors inside each mode.
In fact, this is a particular instance of the renowned “duality” (in this case)
between individuals and organizations (Simmel, 1955, Breiger, 1974). Moreover,
here, we will be concerned only with the particular case of interorganizational networks, which are formed by patterns of participation of individuals (activists) in
various (social movement) political organizations. In this setting, by being members of (or affiliated to) different social movement organizations, activists create
and sustain the linkages of the corresponding interorganizational networks (Diani,
2004), through which personal identities of activists and collective identities of social movements may be structurally constituted and, at the same, these identities
may be culturally weaving the complex entanglements of activists and social movements. For instance, such processes were activated by those that Hedstr¨om et al. ,
2000, have called “socialist agitators,” i.e., traveling activists operating as human
bridges between movement milieus in different localities.
When it comes to the coupling of structure and culture in interorganizational
networks, there has not yet been attained any considerable methodological or formal advancement for the case of two-mode networks similar to the analytical and
statistical studies on homophily and focus constraint applied to the case of onemode networks (that were cited above). However, it is our aim to discuss here
some first steps on a graph–theoretic formalization of how culture matters in the
analysis of certain structural aspects of interorganizational networks (related to
community categorization, as we will explain in the next section).
To do so, we need first to set out with an elementary notation and certain distinctive terminology of some rather trivial concepts in interorganizational networks
that we will be using here – as a matter of fact, we are going to use almost the
same notation with Ronald Breiger (1974). So, let us assume that we have collected a set of empirical data through a questionnaire (or interviews etc.), in which
a number of activists – sometimes referred as participants – report their participation (or involvement or affiliation or membership etc.) to a number of (social
movement) organizations. In particular, we denote by G the set of all organizations
included in the data, say, they are totally g organizations (i.e., |G| = g) and we
denote by P the set of all activists, who are distributed (while overlapping) among
these organizations, say, they are totally p activists (social movement participants).
Hence, such data correspond to a matrix A = A(P, G) of order p × g (i.e., a matrix
composed of p rows and g columns), which may be represented by a two-mode network (or a bipartite graph) of activists and organizations, denoted as H = H(P, G).
Using the generic symbols Oi , Oj , Ok , . . . for organizations, given two arbitrary
organizations, Oi and Oj , let us denote by Bij the set of all participants, who are
4
M.A. BOUDOURIDES
affiliated to both organizations Oi and Oj . Note that Bij is a subset of P (formally,
one writes Bij ∈ 2P , the power set of P ) and, denoting
P by |Bij | the cardinality (i.e.,
the count of membership) of these (sub)sets, then i,j∈P |Bij | ≥ p = |P | (where
| | denotes set cardinality), exactly because members (participants) are overlapping
among organizations. From now on, we will be referring to the set Bij as the “band
of overlapping members” in the two organizations Oi and Oj . Of course, the name
makes sense only as far as the set is non-empty, Bij 6= ∅, i.e., as far as there exists
at least one activist involved with both organizations. If we find all possible (nonempty) bands of membership overlaps throughout our data, then the fundamental
postulate of the persons–groups duality theory says that this set may assess the
existence of indirect relationships (ties) among organizations. The idea of how this
is done is the following:
• If, for two organizations Oi and Oj , it happens that Bij = ∅ (i.e., if there
exist no overlapping members to both organizations), then the organizations Oi and Oj are not linked to each other. Thus, the necessary condition
for two organizations to be linked together in the interorganizational network is that the corresponding (to these organizations) band of overlapping
members is a nonempty set.
• If the cardinality (i.e., the count of membership) of a band Bij of overlapping members to organizations Oi and Oj is larger than the cardinality of
a band Bkl of overlapping members to two other organizations Ok and Ol ,
then the former organizations (Oi and Oj ) possess an intenser tie than the
latter (Ok and Ol ).
Therefore, as far as one admits that the totality of bands of membership overlaps
may assess the existence of all possible interorganizational ties and, furthermore,
that they may measure the intensity of these ties, then one obtains the projection
of the two-mode network H = H(P, G) of activists and organizations on a onemode network G = G(G) of organizations, which is the interorganizational network
we intend to study. Actually, G is a weighted network, meaning that its links
possess a weight, which is expressed by the numerical cardinality (i.e., the count of
membership) of the occurring bands of overlapping membership. For instance, the
weight of the (indirect) link among organizations Oi and Oj is equal to |Bij |. Now,
noting that (for i = j) |Bii | is just the (total) membership of organization Oi , it
|B |
turns out that the normalized weight min{|Biiij|,|Bjj |} may give rise to an indicator of
organizational structural embeddedness, according to Gulati & Gargiulo (1999) (cf.,
Andretta et al. , 2009). In any case, neglecting the aforementioned normalization,
the adjacency matrix of the one-mode interorganizational network G is a symmetric
matrix A = A(G) of order g × g, where the (i, j) entry of A is equal to the weight
|Bij |. Formally, ignoring diagonal terms, this (adjacency) matrix is given by the
following matrix product (Breiger, 1974):
A(G) = A(P, G)T A(P, G),
where A(P, G)T is the transpose matrix of the two-mode matrix A(P, G) – typically,
extracted from the data of the questionnaire.
At this point, let us make a rather evident remark on how an interorganizational
network differs from a personal network (say, a network among persons being tied
by such relationships as friendship, support, affect etc.). In a personal network ties
CULTURE MATTERS IN COMMUNITY INTERORGANIZATIONAL STRUCTURE
5
are sustained by a certain direct action that an actor (person) may develop with
regards to another actor (person). However, in an interorganizational network,
collective actors (organizations) are not seen as directly related to each other, but
only indirectly through the intermediation of individual actors (participants to organizations). In other words, a tie between two organizations is only sustained
indirectly by those participants who act as a kind of duplex proxies (or brokers) of
these organizations. In our terminology, these are the participants who constitute
the discrete bands of overlapping membership and, in this oblique way, they may
bind organizations together. Of course, such an indirect linkage of organizations
by intermediate bands of membership overlaps is only speculated to represent and
surrogate the actual interorganizational (social) relationships.
Now, a direct way to see the traces of culture on interorganizational networks
is by analyzing how overlaps in participants’ membership are differentiated in the
presence of cultural attributes or attitudes that actors (participants) possess. The
underlining idea has been already explored by Cornwell & Harrison (2004), who,
in their study of the interorganizational network among unions and voluntary associations, have argued that: “Analyzing how members of various sociodemographic
groups tie unions to other organizations may provide additional insight into the
basis of unions’ embeddedness” (p. 869). For this purpose, in order to analyze
the influence of certain sociodemographic attributes (such as sex, race, class and
religion) on the way unions are embedded in a broader organizational ecology, Benjamin Cornwell and Jill Ann Harrison first isolated the patterns of membership
overlaps within each attribute and then they tested which attributes were more
effective in tying unions into the overall interorganizational network.
Thus, a graph-theoretic formulation of how to analyze the effect of cultural attributes on an interorganizational network structure might proceed as follows. Let
Q be a (finite) set of attributes or attitudes possessed by activists (participants)
in P reporting their affiliation to multiple (social movement) organizations in G,
the way this information was encoded in the two-mode data matrix A extracted
from an empirical survey (questionnaire etc.). Of course, each activist (participant)
might declare possession of multiple attributes or attitudes from Q. In any case,
(α)
for each attribute α ∈ Q, let us denote by Bij the set of all participants, who
are simultaneously affiliated to two organizations Oi and Oj , while, at the same
time, they are both possessing the attribute α. (Practically speaking, to say that
a participant “possesses an attribute or attitude,” we mean that this participant
has responded affirmatively to the question of the questionnaire that corresponds to
this attribute or attitude.) Similarly to the name we have given for the set Bij , let’s
(α)
call Bij “band of overlapping members possessing the attribute or attitude α” or,
(α)
shortly, just “band of overlapping α-members.” Trivially, Bij is a subset of P (α) ,
P
(α)
i.e., of the set of all participants possessing attribute α, and α∈Q |Bij | ≥ |Bij |,
exactly because members (participants) may overlap among attributes. Therefore,
similarly as before, one obtains the projection of any one of the two-mode networks
H(α) = H(α) (P (α) , G) of α-activists and organizations on the one-mode interorganizational and inter–attributional networks G (α) = G (α) (G). Clearly, the G (α) ’s are
again represented by weighted graphs with symmetric adjacency matrices equal to
6
M.A. BOUDOURIDES
the product {A(α) }T A(α) , where A(α) is the two-mode data matrix restricted just
to α-participants.
Then, in order to elucidate the idea that culture may influence structure in interorganizational networks, it is very instructive to adopt a metaphorical scheme
of immersing ourselves into a rather vivid perspective. In such a pictorial account,
cultural attributes and actors’ attitudes may be graphically conceived as distinct
‘colors,’ which are intertwining a ‘colorful’ patchwork of various partitions of bands
of overlapping membership that constitute an interorganizational network. However, when one focuses exclusively on a certain ‘color’ in these partitions, i.e., when
bands of membership overlaps are restricted only on actors (participants) satisfying
a single cultural attribute or attitude, then the structure of the interorganizational
network might appear different than in patterns, in which ‘colors’ were either erased
or superimposed. Hence, from a graphical point of view, to decipher the volatile
traces of culture amounts to be able to discern the added value that a cultural ‘coloring’ might give to the ‘colorless’ bulk of an interorganizational network, in which
cultural attributes and actors’ attitudes were ignored or compensated. Needless
to say that, since bands of co-participants are typically overlapping (in the sense
that, typically, there exist participants affiliated to multiple organizations, where
‘multiple’ means larger than two), the decomposition of the overall interorganizational network into its ‘colored’ components is not additive or linear. Therefore, it
is possible that certain structural characteristics may become more visible, if not
revealed more clearly, only when an interorganizational network is ‘colored’ by a
certain attribute or attitude, while some other structural features might be dumped,
when the network appears deprived of the cultural qualities of actors. Having said
that, we should not forget that, since an interorganizational network is represented
by a ‘weighted’ graph, the overall effect of culture on structure depends on how
the existing cultural attributes and attitudes of actors are distributed over the interorganizational network ties. As a consequence, the cultural effect on structure is
more salient for certain structural characteristics of the interorganizational network
than others. In the next section, we are going to discuss an example of a structural
feature, which is not only quite sensitive to cultural variation, but, moreover, it
recapitulates substantively the celebrated duality between structure and culture.
3. Communities and Culture
We claim that an appropriate structural feature, on which cultural effects may
plainly imprint, is what is usually named “community structure.” The defining
idea is very simple: in a social network, a community partitioning or structure consists of distinct communities, which are identified with cohesive groups of tight-knit
actors (Wasserman & Faust, 1994, p. 249). In other words, actors inside a community should be relatively more densely connected to each other than how (relatively
sparsely) actors among different communities were connected.
Before entering into a technical discussion about the mathematical and computational operationalization of a community partition in a social networks and, in
the sequel, about the cultural effect on interorganizational network structures, it
is very timely to examine what the concept of community really signifies in social
CULTURE MATTERS IN COMMUNITY INTERORGANIZATIONAL STRUCTURE
7
theory. Undoubtedly, to say to which community one belongs is a peculiar type
of a categorization. It is rather a group or collective categorization (a social categorization, on a more comprehensive level). In fact, a community categorization
has nothing to do with what a secluded individual says or imagines of oneself or
even of one’s group peers. To categorize the community to which one belongs is
not an issue of an enunciated self-identification, even when an individual speculates
on existing affinities with other individuals. For a community does not consist of a
mere aggregation of persons, conceived as self-reliant subjects, who cogitate about
their destiny or recognize themselves in arbitrary or sempiternal terms outside the
reality of their social worlds, their everyday practices and life. On the contrary,
to categorize one’s community is an issue of what one does in practical contexts,
the distinctive ways one acts in mundane situations, how one interacts with or with
regards to others. Even a person’s autonomy (Castoriadis, 1998) or a personal identity, i.e., whatever distinguishes a person from others, are singular characteristics,
which develop through a bundle of inter-personal shared significations, embedded
in the relational footing of the multiplex ties, structural and cultural, that persons
sustain with others in the settings of their common social spaces. Not only “communities can be seen as built out of identities rather than persons,” as Harrison White
(2008a, p.157) has argued, following the historical investigations of Allan Silver
(1989, 1990), but, in addition, “the relationship between community as a complex
of social relationships and community as a complex of ideas and sentiments” needs
further exploration, as Craig Calhoun (2009, p. 110) was calling forth already some
time ago.
Although a complete analytical exploration of this relationship is still an unsolved problem, from a formal methodological point of view, it is also challenging
to try to track down a graph-theoretic approach, in which structural communities
would be conceived formally as group categorizations. Actually, this is what we
are trying to do here. However, in this perspective, communities should not be
perceived immobilized and framed within rigid absolute boundaries. Of course,
typically, there exist relational boundaries, structural cracks or interactional clefts
among communities of a network partition. But these relational boundaries may
interact with other existing symbolic or social boundaries inside the social spaces
where actors live (Lamont & Moln´ar, 2002). And the emerging outcome of the
structure-culture coupling is a process of superlative complexity, like a process of
social annealing with unanticipated and contingent consequences. The effect of
dipping structure inside the contested or troubled waters of an encompassing fluid
cultural milieu might resemble a sort of an inadvertent refraction or a mutated recombination of the preceding ordered pattern of structural categorizations. Because
culture is grounded on so many contrasting or conforming distinctions, contested
values or tastes or opinions, symbolic or linguistic significations, institutional or
pilot arrangements, consensual or dissenting decisions or positions, social or economical inequalities or opportunities, just or unequal distribution of resources etc.
After the previous speculative remarks, let us give a brief description of a formal–
mathematical approach aiming to detect the community structure of a social network. We will follow a technique based on the maximization of a function called
modularity. We have chosen this method, because it is conceptually simple and, for
8
M.A. BOUDOURIDES
small graphs, one could even detect communities by a hands-on computation (e.g.,
for graphs with less than 5 or 6 vertices), although there are many other methods
for the graph-theoretic community detection (cf., Fortunato, 2010). Just to mention
two other, purely sociological, approaches for community (otherwise called cohesive
or social groups) analysis, we may refer to the works of Frank (1995) and Moody
& White (2003).
To start with a short presentation of the modularity maximization technique of
community detection, let G be a social network, which is represented, in general,
by a weighted (undirected simple) graph on a set of vertices V and let n be the
total number of vertices and m the total number of edges of this graph (counting
multiplicity, since the graph is weighted). Furthermore, let us denote by A = {Aij }
the adjacency matrix of the weighted graph (which means that A is a symmetric matrix of order n × n with its entries being non-negative integers such that
P
n
i,j=1 Aij = 2m). (Of course, in the previous notation, G might be anyone of the
projected one-mode interorganizational networks G(G) or G α (G), for any attribute
or attitude α of actors in P .) Furthermore, let C = C(G) = {C1 , C2 , . . . , Cc }
be a partition of the set of vertices V of the graph (representing) G into c subsets Ck . (Technically, such a family of subsets of V forms a “partition,” whenever
S
c
k=1 Ck = V and Ck1 ∩ Ck2 = ∅, for any k1 6= k2 in {1, 2, . . . , c}.) Then a partition C is called community structure of G and each Ck is called a community, if a
certain benefit function is maximized over this partition C. The most commonly
employed benefit function in community partitioning is the following function Q,
called modularity, which was defined by Newman & Girvan (2004) as:
Q = (fraction of links within communities)
– (expected fraction of such links).
In the so-called null model, the expected fraction above is calculated on the basis
of the hypothetical existence of a random graph, which would preserve the same
degree distribution with the examined graph G. Thus, the exact expression of
modularity becomes (Fortunato, 2010):
c h
X
lk dk 2 i
−
,
Q=
m
2m
k=1
where lk is the total number of edges inside community Ck and dk is the sum of
degrees of all vertices in Ck (in both cases, counting multiplicity of edges, since the
graph was assumed weighted). Thus, to obtain the community partitioning one
has just to search for that partition, which maximizes the above modularity function Q. Note that this optimization problem has been proven to be NP-complete
(Brandes et al. , 2008) and, therefore, only approximate optimization techniques,
such as greedy algorithms, simulated annealing, extremal optimization, expectation
maximization, spectral methods etc. can be practically useful.
Returning to our formal analysis of interorganizational networks, when actors
(participants) may exhibit a number of cultural attributes or attitudes, next to the
community partition C = C(G) = {C1 , C2 , . . . , Cc } of the “colorless” interorganizational network G = G(G) (which is the one-mode projection of the two-mode
network H = H(P, G) of activists and organizations), we have a number of “colored” interorganizational networks too, arising from the supplementary information
CULTURE MATTERS IN COMMUNITY INTERORGANIZATIONAL STRUCTURE
9
that we might have in our data (questionnaire etc.), according to which each activist (participant) may possess also a number of cultural–“colored” attributes or
attitudes. Let us denote by Q = {α1 , α2 , . . . , αq } the family of all cultural attributes or attitudes (q in number) that actors (participants) may possess in the
data. This means that, corresponding to each attribute or attitude αl ∈ Q, we have
a “colored” interorganizational network G αl = G αl (G) (as the one-mode projection
of the two-mode network Hαl = Hαl (P αl , G) of αl -activists and organizations,
which is extracted from the original data). Moreover, corresponding to each one of
these “colored” interorganizational networks, we may find a community partition
αl
Cαl = Cαl (G αl ) = {C1αl , C2αl , . . . , Cc(l)
}. Note that the cardinality c(l) = |Cαl | of
any “colored” community partition depends on the attribute or attitude αl .
Needless to say that the constituency of any “colored” community Ckαl depends
on the attribute or attitude αl too. Therefore, and this is very important, “colored” communities may overlap, meaning that an organization may be situated
inside different “colored” communities, for different “colors,” i.e., with regards to
different cultural attributes or attitudes that actors possess. In other words, although community constituencies (in organizations) may change across “colors,”
one can keep account for these changes in order to be able to keep track of the observed recombinations of the community categorizations, when culture enters the
game. Obviously, as “colors” change, if two organizations tend to be inregistered
in the same “colored” communities more frequently than two other organizations,
then one may postulate that the former pair of organizations is more tightly-knit
than the latter pair, because their linkage (of belonging to the same community)
tends to persist sustaining even under different “colors” (attributes or attitudes).
In this way, one may talk of organizational meta–communities, which encapsulate
the effect of culture on structure in the observed culturally–driven mutations of
the interorganizational network structures. Actually, these culturally constituted
meta–communities represent the existing interorganizational alliances, boundaries
or cleavages better than what could have been observed solely in the culture–free
interorganizational structure.
Formally, to get the combined cultural–structural meta–community categorization, all we have to do is to form the matrix M = M (G, C) (where C = {Cα1 , Cα2 ,
. . . , Cαq }) with g rows (the surveyed organizations) and q columns (the vector of
community partitions with regards all the surveyed cultural attributes or actors’
attitudes), the (i, j) entry of which is the index of the community to which the
i-th organization Oi belongs in the community partition Cαj with regards to the
j-th attribute or attitude αj . Of course, M is not a binary matrix, but it can
be converted to a binary matrix (still denoted as M ), if we recode its columns
αl
}, for
by expanding to the corresponding community structure {C1αl , C2αl , . . . , Cc(l)
eachP
cultural attribute al , for l = 1, . . . , q. Thus, the binary matrix M is of order
q
g × l=1 c(l). Then, apparently, M represents a two-mode network (a bipartite
graph), denoted as M, among organizations and culturally–“colored” communities.
The corresponding projection of M on just organizations would be an one-mode
interorganizational network N with adjacency matrix equal to the matrix product
M M T (i.e., a weighted graph). We call N “community multiplexity counting interorganizational network,” because the (i, j) entry of its adjacency matrix equals to
10
M.A. BOUDOURIDES
the number of different culturally–“colored” communities, to which the two organizations Oi and Oj belong. Thus, determining the community partition of the latter
network N , we find the structure of meta–communities C = C(G) = {C1 , C2 , . . . , Cµ }.
In the next sections, we are going to give two examples of the previous formal
methodology and the corresponding computations. These examples intend to manifest the added value of incorporating cultural attributes and individual actors’ attitudes in the study of organizational networks. The first example is an artificial one
and it concerns a situation, in which there might be no visible community structure in an interorganizational network. However, communities may emerge only
as far as one starts “coloring” the patterns of the observed multi-organizational
overlaps by considering the cultural attributes that participants may exhibit. The
second example is derived from certain concrete empirical data (the IPPS survey,
a typical survey about participation in a recent global mobilization) and what it
manifests is the fact that the consideration of the activists’ attitudes towards issues
of peace and war (together with their membership overlaps) might enhance the
evidence about the emergence of a clinging interorganizational alignment (what we
call here “meta–communities”) among participating social movement organizations,
with which surveyed activists are affiliated, due to the existing patterns of shared
common attitudes with regards to the salient issues that frame the mobilization, in
which activists are participating.
4. An Artificial Example
First, let us consider the interorganizational network among four organizations, A,
B, C and D, which is represented in the following weighted graph:
D
A
4
8
10
B
8
C
Figure 1. A “colorless” (i.e., attribute–free) interorganizational network.
Here, one can prove (formally) that, for any partition of the “colorless” graph,
there exists no positive modularity function, which is maximized over that partition. This means, technically speaking, that the “colorless” interorganizational
network of the example does not possess any community structure (in the formal
way it was defined previously).
CULTURE MATTERS IN COMMUNITY INTERORGANIZATIONAL STRUCTURE
11
Furthermore, let us assume that the participants to the above four organizations
exhibit three attributes, denoted by the three “colors,” red, blue and green. In particular, let us suppose that the patterns of participants’ possession of these four attributes are such that three “colored” interorganizational networks are formed. The
“red” interorganizational network is represented in the following weighted graph:
D
A
3
6
2
B
C
2
Figure 2. The “red” interorganizational network.
Now, it is not difficult to compute that the modularity maximizing community
partition of the “red” network admits two communities of organizations: {A, B}
and {C, D}.
Next, the “blue” interorganizational network is represented in the following
weighted graph:
D
A
3
6
B
C
Figure 3. The “blue” interorganizational network.
Again, it is trivial to compute that the modularity maximizing community partition of the “blue” network admits two communities of organizations: {A, C, D}
and {B}.
Finally, the “green” interorganizational network is represented by the following
weighted graph:
12
M.A. BOUDOURIDES
D
A
1
8
2
B
C
2
Figure 4. The “green” interorganizational network.
Here too, one can easily compute that the modularity maximizing community partition of the “green” network admits three communities of organizations:
{A, B}, {C} and {D}. (Note that, in the last two “colored” partitions, some communities are singletons, i.e., they are composed of single organizations.)
We have to admit that, in the previous community detection computations, we
have not been taking into consideration the presence of self–loops of nodes. Of
course, this is just a technical omission that could be definitely surmounted in a
more comprehensive and exhaustive computation, in which one would have to start
from the full two-mode data of all participants affiliated to organizations and exhibiting a number of attributes. However, we have made this omission just in order
to be able to illustrate the peculiar situation of a community–free “colorless” network coupled with a number of community–available “colored” networks.
Concisely, the distribution of the four organizations in the corresponding “colored” communities that we have found is given by the following table (where cell
entries denote the indices of the corresponding communities in the columns):
Organizations
A
B
C
D
“Colorless”
–
–
–
–
Communities
“Red” “Blue”
1
1
1
2
2
1
2
1
“Green”
1
1
2
3
From the above table, one can easily derive what, in the previous section, was
called “community multiplexity counting interorganizational network” N and find
that the adjacency matrix of N is the following:

0
2

1
1
2
0
0
0
1
0
0
2

1
0

2
0
CULTURE MATTERS IN COMMUNITY INTERORGANIZATIONAL STRUCTURE
13
Therefore, in the interorganizational network of this example, two meta–communities emerge, {A, B} and {C, D}, which coincide with the “red” communities, although in the overall network (containing all “colors”) no communities could be
discerned. Therefore, in certain structurally ambiguous situations that one may
encounter in the context of concrete empirical cases, when there appears no community structure to be disclosed by mere structural considerations, then it is substantial that one might focus on any available “colored” cultural attributes in order
to be able to discern (at least) some elements of a (possibly partial) community
categorization, something which would have been lost in the absence of any focalization on culture.
5. The Case of Italy in the IPPS Survey
Let us now come to our second example. This is an interorganizational network,
which was extracted from the empirical data (two-mode networks of organizational
overlaps in participation and attitudinal possession) of the IPPS survey. In fact,
on February 15, 2003, mass protests against the imminent (that period) war in
Iraq took place throughout the world, in which more than seven million people in
more than 300 cities all over the globe had participated. This globally sustained
mobilization was one of the largest peace protests since the Vietnam War on one
single day.1 On this occasion, an international team of social movement scholars
set up a project, called International Peace Protest Survey or IPPS in abbreviation
(2003-4), which was coordinated by Stefaan Walgrave of the University of Antwerp
in Belgium (Walgrave & Verhulst, 2009, Diani, 2009).2 The aim of this project was
to study the demonstrations of the global protest event of February 15, 2003, and
compare its social movement dynamics in 8 countries: UK, Italy, the Netherlands,
Switzerland, USA, Spain, Germany and Belgium. For this purpose, over 10,000
questionnaires were totally distributed and about 6,000 completed questionnaires
were sent back, which makes the successful response rate quite satisfactory (well
above 50%).
In particular, in the IPPS survey, participating activists (with varying numbers
in each country) were asked to declare their affiliation or involvement with the
following 16 (types of) organizations:
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Church
Anti-Racist
Student
Labor Union – Professional
Political Party
Women
Sports – Recreation
Environmental
(9)
(10)
(11)
(12)
(13)
(14)
(15)
(16)
Art, Music & Education
Neighborhood
Charitable
Anti-Globalist
Third World
Human Rights
Peace
Other
1For more descriptive information see the Wikipedia article http://en.wikipedia.org/wiki/
February_15,_2003_anti-war_protest.
2
The survey data bases are accessible at http://webh01.ua.ac.be/m2p/index.php?page=
projects&page2=pproject&id=11.
14
M.A. BOUDOURIDES
Furthermore, respondents were asked to answer a number of questions, in which
they would express their own opinions and feelings about the meaning, the significance, the reasons, the implications etc. of the war. In particular, activists
participating in this event were asked to disclose their positions on the following 10
war–related attitudes:
(1) USA Crusade against Islam
(2) Anti-Dictatorial Regime War
(3) UN Security Council Authorized
War
(4) War for Oil
(5) Racist War
(6) Iraqi Threat to World Peace
(7) Always Wrong War
(8) War to Overthrow the Iraqi
Regime
(9) Feelings against Neoliberal
Globalization
(10) Governmental Dissatisfaction
Here, we are going to discuss just the findings of the analysis of the IPPS survey
data from a single country, Italy, where 972 activists, participating in the protest
events of February 15, 2003, have responded to the questionnaire. In the following
table, we may see what we have found (following the previously discussed analysis)
on the distribution of the 16 organizations in various communities formed by the
possession of the 10 war–related attitudes by the participating activists. Again,
cell entries denote the indices of the corresponding communities in the columns.
However, now, the first column corresponds to the case of the “colorless” interorganizational network, in which activists’ attitudes were completely ignored, while the
last column corresponds to the case of the entirely “colorful” interorganizational
network, in which all activists’ attitudes were taken into account (incorporated into
the community multiplexity counting interorganizational network N , the meta–
community structure of which is given in the last column).
No
attitudes
Church
Anti-Racist
Student
Labor Union – Prof.
Political Party
Women
Sport – Recr.
Environmental
Art, Music & Edu.
Neighborhood
Charitable
Anti-Globalist
Third World
Human Rights
Peace
Other
8
3
4
1
1
5
7
2
7
9
6
3
3
3
3
6
USA
Crusade
against
Islam
7
6
6
4
4
1
5
6
5
3
2
8
8
8
8
2
AntiDictatorial
Regime
War
4
1
3
2
2
1
3
1
3
3
1
1
1
1
1
1
UN
War
Sefor
cuOil
rity
Council
Authorized
War
1
4
3
3
3
2
2
4
2
3
1
4
4
4
4
1
8
2
3
1
1
4
6
5
6
9
7
2
2
2
2
7
Racist
War
6
2
4
7
7
5
6
2
6
1
6
3
3
3
3
5
Iraqi
Threat
to
World
Peace
5
2
5
1
1
2
5
6
5
7
3
4
4
4
4
3
Always
Wrong
War
3
7
1
6
6
8
4
9
4
2
5
7
7
7
7
5
War
to
Overthrow
the
Iraqi
Regime
1
2
1
6
6
2
1
2
1
6
3
4
4
4
4
5
Feelings
against
Neoliberal
Globalization
5
1
1
4
4
1
2
1
2
3
6
1
1
1
1
5
Gov- All
ernattiment- tudes
al
Dissatisfaction
3
3
4
3
4
3
2
3
1
3
1
3
1
1
4
4
2
3
2
1
1
3
2
3
2
1
2
3
3
3
3
2
The following two graphs portray the interorganizational network of activists,
who have participated in the February 15, 2003, protest events in Italy (as they
were surveyed by IPPS). In these graphs, the participating organizations are colored
CULTURE MATTERS IN COMMUNITY INTERORGANIZATIONAL STRUCTURE
15
according to the community or meta–community they belong: on the first graph,
when the activists’ attitudes are ignored in the interorganizational network G, and,
on the second graph, when all these attitudes are cumulated in the community
multiplexity counting interorganizational network N .
Art, Music & Edu.
Charitable
Neighborhood
Student
Environmental
Sport & Recr.
Church
Third World
Human Rights
Anti−Globalist
Political Party
Peace
Women
Other
Anti−Racist
Labor Union & Prof.
Figure 5. The interorganizational network G in Italy ignoring
activists’ attitudes and portraying the corresponding communities
of the participating organizations with different colors.
Political Party
Labor Union & Prof.
Neighborhood
Environmental
Anti−Globalist
Student
Peace
Other
Church
Anti−Racist
Women
Sport & Recr.
Human Rights
Charitable
Third World
Art, Music & Edu.
Figure 6. The community multiplexity counting interorganizational network N in Italy taking into account all activists’ attitudes and portraying the corresponding meta–communities of the
participating organizations with different colors.
16
M.A. BOUDOURIDES
The comparison between communities (or meta–communities) of organizations in
these two networks is fascinating. First of all, the “colored” community categorization (i.e., taking into account activists’ attitudes) is more succinct – composed of 4
meta–communities – than the “colorless” one (in the absence of activists’ attitudes)
– composed of 9 communities. To start with, from the above two interorganizational
graphs, it is evident that there exist two attitudinally derived meta–communities
that almost coincide with two corresponding structural communities. First, we have
a group of five organizations (Anti–Globalist, Peace, Anti–Racist, Human Rights
and Third World), which stick together into the same meta–community (and, very
frequently, they appear into the same “colored” communities too), but they also
constitute one of the 9 communities of the attitudes–free interorganizational structure. Note that the full constituency of this meta–community (in Italy) includes
two more organizations, Environmental and Women, which were structurally classified in communities composed of single organizations (singletons). Incidentally,
based on purely structural considerations, Mario Diani (2009) has classified the
meta–cluster of these seven organizations as the group of “peace protest communities.” Secondly, there exists a group of two organizations, Political Party and Labor
Union – Professional, which, just by themselves, compose a single structural community, but in the inter–attitudinal network they partake with a third organization,
Neighborhood, in the constitution of one of the existing three meta–communities.
In these two cases, one could only speculate that the effect of the attitudinally expressed culture into the emergence of the structural concrescence of the constituent
organizations is so profound that it cannot be singled out from the purely structural
interorganizational entanglement of the observed overlapping memberships.
Finally, coming back to the Italian protest network, there are cases manifesting
how the influence of culture (through the possession of war–related attitudes by
protesters) may facilitate the emergence of some coherent interorganizational meta–
communities, which in the “colorless” network were fragmented and dispersed in
disparate structural communities. A very characteristic example is the appearance
of a meta–community composed of five organizations (in Italy), resembling what
in the terminology of Mario Diani (2009) was called “peace associations”: Student, Church, Sports–Recreation, Charitable and Art, Music & Education (a sixth
“organization” in this meta-community is the residue “Other”). In the structural
community categorization, only two of these six organizations were clustered in the
same community (Sports–Recreation and Art, Music & Education), while the remaining three were structurally classified in communities composed of singletons.
We conclude this section by showing the total meta–community multiplexity
counting network, which is formed by gluing together the outcomes of the analyses
of all the eight countries partaking in the IPPS survey (similar to the Italian case
that we have presented above).
Now, the emerging meta–communities at the international level are just two.
One is the meta–community of the six organizations: Anti–Globalist, Peace, Anti–
Racist, Human Rights, Third World and Women. The other is the meta–community
of the remaining nine organizations (exempting the residual Others): Environmental, Student, Church, Sports–Recreation, Charitable, Art, Music & Education,
CULTURE MATTERS IN COMMUNITY INTERORGANIZATIONAL STRUCTURE
17
Anti−Globalist
Neighborhood
Anti−Racist
Peace
Women
Human Rights
Political Party
Sport & Recr.
Labor Union & Prof.
Other
Environmental
Third World
Art, Music & Edu.
Student
Church
Charitable
Figure 7. The meta–community multiplexity counting interorganizational network in the eight countries of the IPPS survey, taking
into account the attitudes of all activists’ (from all countries) and
portraying the corresponding meta–communities of the participating organizations with different colors.
Political Party, Labor Union & Professional and Neighborhood. This interorganizational cleavage is daringly reminiscent of Mario Diani’s (2009) contradistinction
between “protest communities” and “associations.” However, the difference is that
now the cleavage has been derived analytically (by a formal methodology on a combined structural–cultural community categorization) than been speculated a priori
in the context of a plain structural analysis, which would have discarded any available cultural–attitudinal evidence in the survey. As it is sometimes said, one needs
to be cautious not to throw out the baby with the bath water!
6. Discussion
Let us start by commenting on a seemingly rather convincing negative assessment
of the utility of cultural availability and significance to studies of social movement
participation that has been made a long time ago by Doug McAdam (1986) and
it seems to have discouraged the attention of many social movement scholars on a
possible investigation of the structure–culture duality in the context of participation
studies. This is the argument of the “attitudinal affinity” that McAdam exposes
very clearly when he argues (p. 65): “The argument is that structural availability is
more important than attitudinal affinity in accounting for differential involvement
in movement activity. Ideological disposition toward participation matters little if
the individual lacks the structural contact to ‘pull’ him or her into protest activity.”
McAdam is right, as many recent empirical studies have been demonstrating, that
the decisive and sometimes even predictive role in encouraging activism or recruitment to at least high–risk forms of social movement participation is rather structural
than cultural or attitudinal. However, when it comes to explaining the emergence
of concrete structural patterns of certain macro–level interorganizational features
18
M.A. BOUDOURIDES
(as, for example, the diversity in the observed alignments in particular community
categorizations), the structural argument is completely mute. It might be true that
an individual is not pulled into protest activity because of ideological dispositions or
cultural–attitudinal affinities, but because of the structural interlocking within certain constricting multi–organizational membership overlaps. Nevertheless, a mere
count of the constituencies of the structural interorganizational interlockings does
not suffice to justify why and to account how such alignments or other particularized patterns come to exist.
Yes, it is indeed reasonable to accept the structural rationale that two organizations, say, X and Y , should be inter–connected as far as they do accommodate
some considerable number of overlapping members. Moreover, under the structural
rationale, we can draw a map (a graph of the corresponding interorganizational
network) portraying the entanglements among the relational positions of all observed organizations. But this graph is typically highly dense (because of the often
high rates of observed co-membership overlaps) and it might be the case that the
two organizations X and Y were still connected in it through a moderate rate of
co-memberships. However, if it also turns out that X and Y are situated within different parts of this interorganizational network (for instance, they may be aligned
within the camps of two different structural communities) or they might exhibit
some diverse structural features (like centralities, core–periphery location, equivalences etc.), then how can the purely structural argument explain or construe such
circumstances? No, it can’t. This is why the structurally unwanted patterns of
cultural or attitudinal affinities need to be called back in order to elucidate such
structural differentiations by inspecting their nuances with a palette of “colors”
(i.e., cultural distinctions or significations), which might revive the existing diversity that the structural rationale has cached away.
To give a simpler example, say, the overlapping members between two pairs of
organizations are 50 in both cases. But this is just a number and the structural
rationale is basing all its explanatory power on the number 50 in relation to other
such numbers counting the existing membership overlaps among all other pairs of
organizations. Nevertheless, one does not need any sophisticated philosophical argument to understand that, in reality, there are qualities hidden beyond the surface
of quantities, which are counting various entities in the world. It is sufficient to
argue that the 50 participants interlocking one pair of organizations may be either
(i) the same or (ii) completely different or (iii) overlapping with the 50 participants interlocking another pair of organizations. Of course, it is not a matter of
a strict identification of the names of participants, since most of social movement
surveys are necessarily anonymous (or they have to be anonymized, when their
findings are disseminated in public). However, participants possess qualities and
often they are willing to disclose them through the responses they give in social
surveys. This is the case of the cultural or attitudinal characteristics that participants might be asked in the survey to expose. Then, in the light of certain
explicitly expressible cultural diversities, the two quantitative interlocks of a mere
count of 50 co-participants might be decomposed in two different arrays of participants endowed with different cultural or attitudinal characteristics. Actually, this
is the behavioral evidence needed to explain the emergence of macro–patterns, like
CULTURE MATTERS IN COMMUNITY INTERORGANIZATIONAL STRUCTURE
19
community categorizations or alignments or boundaries among different positional
locations, in a structurally composed interorganizational network.
That’s why culture matters in studies of structure. To grasp the cultural significance to structure in full–fledged terms presumes that one has to surpass an old
methodological bias about the marginal role that actors’ attitudes might play in
structural analyses. However, it was from the 1980s that Bonnie Erickson has been
shouting loudly that the right structural way to comprehend how attitudes really
work is by considering that: “(a) natural units of analysis for attitudes are not isolated individuals but social networks and (b) viable subjects for explanation are not
individual attitudes, but degrees of attitude agreement among individuals in given
structural situations” (Erickson, 1988, p. 99). This is exactly what the proposed
formal methodology of the culture–structure duality is aspiring to do. It is not a
matter of aggregating individual actors’ attitudes in a structural group that they
may form – one knows this very well, if one understands the problem with the ecological fallacy. On the contrary, it is a matter of forming the inter-attitudinal network
by aggregating in community categorizations the cultural patterns of “degrees of
attitude agreement among individuals” and, subsequently, juxtaposing them with
the structural pattern of membership overlaps and interlocks in the interorganizational network. This is exactly the approach we have been following here.
Now, coming back to a discussion of some more formal aspects of the proposed
methodology of an interorganizational network analysis based on the structure–
culture duality, let us first remark on something that one might have already
noticed in the previous IPPS example. The cardinality of the culturally–driven
meta–communities appeared there to be much lower than the cardinality of the
purely structural communities. However, this is completely consistent with the fact
that the number of links in the interorganizational network G (i.e., the number
of overlapping memberships of participants in organizations) is much higher than
the number of links in the community multiplexity counting network N (i.e., the
number of overlapping memberships of organizations in organizational communities, which can be at most as many as the number of the reported attitudes, i.e.,
10 in that example). Unless, of course, one was analyzing empirical data containing a very large number of cultural attributes or actors’ attitudes and a relatively
smaller number of organizations (or events etc.), to which actors could have been
interlocked. This was not the case with the IPPS data: a total number of a few
thousands of activists (in eight countries) generating a very large number of tightknit overlapping members of totally 16 organizations, while the surveyed attitudes
were just 10. But, as said, there might be empirical data, in which the numbers of
organizational affiliations could be much smaller than the numbers of attributional
(or attitudinal) affiliations. Perhaps, in such a case, it would be advisable to reverse
the order of analysis: one could first start with the inter–cultural structure (forming
the inter–attributional or inter–attitudinal network) and, subsequently, one could
proceed in decomposing the latter structure according to the sparser patterns of
the organizational affiliations. In fact, this reversed analysis would signify a good
grip on the (reversible) duality between structure and culture.
20
M.A. BOUDOURIDES
Furthermore, from the examples and formal network discussion that we have
made, it is clear that, in the context of networks of participation, culture, through
the incorporation of attitudes that activists happen to possess, may play a cementation role in forcing certain organizations to align inside the same meta–communities,
while these organizations, in the absence of cultural attitudes, would appear inexplicably clustering within certain dissimilar structural groups. As a matter of fact,
global cultural identities in modern societies may still play the role of a functional
substitute for local communities, as, for instance, nation and nationalism once used
to provide the integrative cement that made modern communities appear as the
Gemeinschaft (Haas, 1964, p. 465). Actually, in the contemporary world, this is
exactly what one may observe analytically in social networks and, in particular,
with regards to the constricting effect of culture on structure, as it is typically celebrated by the notion of the “global village,” attributed to Marshall McLuhan.
However, the indication that a cultural–attitudinal analysis may sometimes return the same organizational clusters with the ones obtained by a pure structural
analysis should not be misleading. By all means, this does not mean that the
cultural–attitudinal analysis is redundant. In many other cases (as we have already seen in the Italian IPPS network), the cultural–attitudinal analysis may shed
light on circumstances that would have remained impalpable from the structural
point of view. The problem with the structural analysis of interorganizational networks is that, as a network theory, it is extremely coarse (for instance, because of
the oversimplification caused by the reduction–projection of a two-mode network
on an one-mode network, which dispenses with any existing micro–structural information). Therefore, the interorganizational structural analysis remains strictly
descriptive in character and it is incapable of providing the required background
of (say) some precise micro-localities (like local independences or other complex or
fractal topological features), over which one could apply certain powerful recently
developed statistical or computational techniques (like ERGMs, multi–level analyses, small–worlds, power laws etc.) in order to be able to estimate or simulate the
outcome of those random processes or micro–social mechanisms that could predict
more accurately (and formally rigorously) the routes of the dynamic evolution of
structure.
Of course, neither the cultural–attitudinal analysis may do this. However, the
added value of the latter to a plain, if not exiguous, form of structural analysis
is that culture may provide an empirical heuristic for the interpretation of structure, which would have otherwise remained insubstantially “colorless” or insipid.
Such is the lesson of some macro–sociological or practice–oriented institutional approaches built over the fundamental comprehensive idea of the structure–culture
duality (Breiger, 2000, Breiger & Mohr, 2004, Lizardo, 2006). It is essential that
social network analysts become more acclimated to this lesson, because they still
have many contributions to offer from the angle of the methodological clarity and
the rigorous formalism of the network–relational paradigm.
CULTURE MATTERS IN COMMUNITY INTERORGANIZATIONAL STRUCTURE
21
References
Andretta, Massimiliano, Botetzagias, Iosif, Boudourides, Moses, Kioufegi, Olga,
& Yang, Mundo. 2009. Novel Characteristics of the GJMs: A (latent) network
approach. Pages 149–172 of: della Porta, Donatella (ed), Another Europe: Conceptions and Practices of Democracy in the European Social Forums. London:
Routledge. http://nicomedia.math.upatras.gr/sm/Demos2009.pdf.
Brandes, Ulrik, Delling, Daniel, Gaertler, Marco, G¨orke, Robert, Hoefer, Martin,
Nikoloski, Zoran, & Wagner, Dorothea. 2008. On modularity clustering. IEEE
Transactions on Knowledge and Data Engineering, 20, 172–188. http://dx.
doi.org/10.1109/TKDE.2007.190689.
Breiger, Ronald L. 1974. The duality of persons and groups. Social Forces, 53,
181–190. http://www.jstor.org/stable/2576011.
Breiger, Ronald L. 2000. A tool kit for practice theory. Poetics, 27, 91–115.
http://dx.doi.org/10.1016/S0304-422X(99)00026-1.
Breiger, Ronald L., & Mohr, John W. 2004. Institutional logics from the aggregation
of organizational networks: Operational procedures for the analysis of counted
data. Computational & Mathematical Organization Theory, 10, 17–43. http:
//dx.doi.org/10.1023/B:CMOT.0000032578.16511.9d.
Calhoun, Craig J. 2009. Community: Toward a variable conceptualization for
comparative research. Social History, 5, 105–129. http://www.jstor.org/pss/
4284951.
Castoriadis, Cornelius. 1998. The Imaginary Institution of Society. Transl. Kathleen
Blamey. Cambridge: MIT Press.
Cornwell, Benjamin, & Harrison, Jill Ann. 2004. Union members and voluntary associations: Membership overlap as a case of organizational embeddedness. American Sociological Review, 69, 862–881. http://dx.doi.org/10.1177/
000312240406900606.
Diani, Mario. 2004. Networks and Participation. Pages 339–359 of: Snow, David A,
Soule, Sarah A., & Kriesi, Hanspeter (eds), The Blackwell Companion to Social
Movements. Oxford: Blackwell.
Diani, Mario. 2009. The structural bases of protest events: Multiple memberships
and civil society networks in the 15 February 2003 anti-war demonstrations. Acta
Sociologica, 52, 63–83. http://dx.doi.org/10.1177/0001699308100634.
Emirbayer, Mustafa. 1997. Manifesto for a Relational Sociology. American Journal
of Sociology, 103(2), 281–317. http://dx.doi.org/10.1086/231209.
Emirbayer, Mustafa, & Goodwin, Jeff. 1994. Network analysis, culture, and the
problem of agency. American Journal of Sociology, 99, 1411–1454. http://www.
jstor.org/stable/2782580.
Emirbayer, Mustafa, & Mische, Ann. 1998. What is agency? American Journal of
Sociology, 103, 962–1023. http://dx.doi.org/10.1086/231294.
Erickson, Bonnie H. 1988. The relational basis of attitudes. Pages 99–121 of: Wellman, Barry, & Berkowitz, S. D. (eds), Social Structures: A Network Approach.
Cambridge: Cambridge University Press.
Feld, Scott L. 2000. The focused organization of social ties. American Journal of
Sociology, 86, 1015–1035. http://www.jstor.org/stable/2778746.
Fortunato, S. 2010. Community detection in graphs. Phys. Rep., 486, 75–174.
http://dx.doi.org/j.physrep.2009.11.002.
22
M.A. BOUDOURIDES
Frank, Kenneth A. 1995. Identifying cohesive subgroups. Social Networks, 17,
27–56. http://dx.doi.org/10.1016/0378-8733(94)00247-8.
Friedland, Roger, & Mohr, John. 2004. The cultural turn in American sociology.
Pages 1–10 of: Friedland, Roger, & Mohr, John (eds), Matters of Culture: Cultural Sociology in Practice. Cambridge: Cambridge University Press.
Gulati, Ranjay, & Gargiulo, Martin. 1999. Where do interorganizational networks
come from? American Journal of Sociology, 104, 1439–1493. http://www.
jstor.org/stable/2990941.
Haas, Ernst B. 1964. Beyond the Nation-State: Functionalism and International
Organization. Stanford: Stanford University Press.
Hedstr¨
om, Peter, Sandell, Rickard, & Stern, Charlotta. 2000. Mesolevel networks
and the diffusion of social movements: The case of the Swedish Social Democratic
Party. American Journal of Sociology, 106, 145–172. http://www.jstor.org/
pss/10.1086/303109.
Lamont, Mich`elle, & Moln´
ar, Vir´ag. 2002. The study of boundaries in the social
sciences. Annual Review of Sociology, 28, 167–195. http://dx.doi.org/10.
1146/annurev.soc.28.110601.141107.
Lazarsfeld, Paul F., & Merton, Robert K. 1964. Friendship as social process: A
substantive and methodological analysis. Pages 18–66 of: Berger, Monroe, Abel,
T., & Page, C. H. (eds), Freedom and Control in Modern Society. New York:
Octagon Books.
Leenders, Roger Th. A. J. 1997. Longitudinal behavior of network structure and
actor attributes: Modeling interdependence of contagion and selection. Pages
165–184 of: Doreian, Patrick, & Stokman, Franz N. (eds), Evolution of Social
Networks. Amsterdam: Gordon & Breachs.
Lizardo, Omar. 2006. How cultural tastes shape personal networks. American Sociological Review, 71, 778–807.
http://dx.doi.org/10.1177/
000312240607100504.
McAdam, Doug. 1986. Recruitment to high–risk activism: The case of freedom
Summer. American Journal of Sociology, 92, 64–90. http://www.jstor.org/
stable/2779717.
McPherson, Miller, Smith-Lovin, Lynn, & Cook, James M. 2001. Birds of a feather:
Homophily in Social Networks. Annual Review of Sociology, 27, 415–444. http:
//dx.doi.org/10.1146/annurev.soc.27.1.415.
Melucci, Alberto. 1989. Nomads of the Present. Philadelphia: Temple University
Press.
Melucci, Alberto. 1996. Challenging Codes. Cambridge: Cambridge University
Press.
Moody, James, & White, Douglas R. 2003. Structural cohesion and embeddedness:
A hierarchical concept of social groups. American Sociological Review, 68, 103–
127. http://dx.doi.org/10.1234/12345678.
M¨
utzel, Sophie. 2009. Networks as culturally constituted processes. Current Sociology, 57(6), 871–887. http://dx.doi.org/10.1177/0011392109342223.
Newman, Mark E.J., & Girvan, Michelle. 2004. Finding and evaluating community
structure in networks. Phys. Rev. E, 69, 026113. http://dx.doi.org/10.1103/
PhysRevE.69.026113.
Robins, Garry, Pattison, Philippa, & Elliott, Peter. 2001a. Network models for
social influence processes. Psychometrika, 66, 161–190. http://dx.doi.org/
CULTURE MATTERS IN COMMUNITY INTERORGANIZATIONAL STRUCTURE
23
10.1007/BF02294834.
Robins, Garry, Elliott, Peter, & Pattison, Philippa. 2001b. Network models for
social selection processes. Social Networks, 23, 1–30. http://dx.doi.org/10.
1016/S0378-8733(01)00029-6.
Robins, Garry, Pattison, Philippa, Kalish, Yuval, & Lusher, Dean. 2007. An introduction to exponential random graph (p∗ ) models for social networks. Social
Networks, 29, 173–191. http://dx.doi.org/10.1016/j.socnet.2006.08.002.
Silver, Allan. 1989. Friendship and trust as moral ideals: An historical approach.
European Journal of Sociology, 30, 274–297. http://dx.doi.org/10.1017/
S0003975600005890.
Silver, Allan. 1990. Friendship in Commercial Society: Eighteenth-Century Social
Theory and Modern Sociology. American Journal of Sociology, 91, 1474–1504.
http://www.jstor.org/stable/2780332.
Simmel, Georg. 1955. Conflict and the Web of Group Affiliations. Transl. Kurt H.
Wolff and Reinhard Bendix. New York: Free Press.
Steglich, Christian, Snijders, Tom A.B., & Pearson, Michael. 2007. Dynamic networks and behavior: Separating selection from influence. Socialogical Methodology, 40, 329–393. http://dx.doi.org/10.1111/j.1467-9531.2010.01225.x.
Walgrave, Stefaan, & Verhulst, Joris. 2009. Government stance and internal diversity of protest: A comparative study of protest against the war in Iraq in eight
countries. Social Forces, 87, 1355–1387. http://dx.doi.org/10.1353/sof.0.
0171.
Wasserman, Stanley, & Faust, Katherine. 1994. Social Network Analysis: Methods
and Applications. 1st edn. Cambridge: Cambridge University Press.
White, Harrison C. 2008a. Identity and Control: How Social Formations Emerge.
2nd edn. Princeton, NJ: Princeton University Press.
White, Harrison C. 2008b. Notes on the constituents of social structure. Soc. Rel.
10 – Spring 1965. Sociologica, 1, 1–15. http://dx.doi.org/10.2383/26576.
Department of Mathematics, University of Patras, Greece
E-mail address: [email protected]